This subproject seeks to evaluate two different treatment approaches to aprosodia, acquired deficits in emotional expression after stroke. The basic premise is that aprosodia may have an expressive component and a receptive component, and therefore, patients may respond differentially to feedback and repetition (targeting the expressive component) and linguistic cues (targeting the receptive component). The investigators will perform fMRI scans on normal subjects as well as patients in the study to determine whether patients with aprosodia have activation patterns different from those normal subjects while producing emotionally intoned sentences, and whether these patterns change in response to therapies.